Qualcomm has recently launched its new Snapdragon 8 Gen 2 platform, and you can read our complete Snapdragon 8 Gen 2 overview to learn the key functionalities and new features.
We also had a chance to run some benchmarks on development handsets Qualcomm and its partners use to build software (see photo above).
Such handsets are typically called MDP (mobile development platform) and are not explicitly designed for benchmarks. Still, based on our experience, they provide a relatable performance proxy for upcoming high-end phones.
We looked at the CPU and Graphics performance using well-known benchmarks such as Geekbench and GFXBench, but other benchmarks results mostly correlate with these two regarding CPU and GPU speed.
We’re trying to see where the new Snapdragon 8 Gen 2 pushes the envelope, as seen by most consumers, and CPU + Graphics are an excellent performance indicator. Should you care about benchmarks? Yes, but only to a point.
The data we’ll publish below includes AI (artificial-related) benchmarks. They are pretty interesting to look at if you’re into this. However, AI benefits are not yet directly reflected by AI benchmarks. While we are interested in such benchmarks, we aren’t using them yet in our mobile phone recommendations.
The Samsung Galaxy S22 Ultra represents last year’s Snapdragon 8 Gen 1 platform which dominated the high-end market. The OnePlus 10T features the Snapdragon 8+ Gen 1, a slightly faster version of the original Snapdragon 8 Gen 1.
The Geekbench 5 single-threaded only uses the fastest CPU core and shows a great deal of performance increase over the last-generation platform.
More importantly, when all cores are pushed to their limit, the new Snapdragon 8 Gen 2 does much better than its predecessor by a good margin, and this should show up in any app which requires massive CPU computing.
During daily usage, such CPU performance translates into faster app initialization or shorter wait times for many different tasks. Many people say the difference is not apparent, but all you need to do is use a one or two-year-old phone to feel it right sway. It’s much easier to adopt new tech.
It’s worth noting that Apple’s A15 Bionic and A16 Bionic remain ahead in this particular benchmark, although the gap is reduced compared to previous years.
The graphics performance of the Snapdragon 8 Gen 2 is also impressive and features a considerable boost in benchmarks. Using the excellent OnePlus 10T as a gaming reference phone, you can see how much faster Qualcomm’s new platform is.
Sure, you might not want to play at 200 FPS, but the chip can handle the most complex games available to mobiles, including some that use advanced features such as raytraced reflections and shadows. That’s a more critical byproduct of higher performance than just higher FPS.
The performance jump is even more significant in the latest graphics benchmarks, such as GFXBench Aztec Ruins (Vulkan/Metal, offscreen).
In graphics benchmarks such as GFXBench, Qualcomm’s Snapdragon 8 Gen 2 dominates its Apple A15 and A16 counterparts, which indicates that Qualcomm’s GPU designs and drivers are incredibly advanced.
That’s also true even when using benchmarks that naturally favor Apple, like GFXBench Manhattan 3.1 OpenGL (android) vs. Metal (iPhone). The Metal graphics API is best compared to Vulkan on Android. It features lower driver CPU overhead and tends to score higher than OpenGL. Yet, Qualcomm still wins.
Regarding neural processing or artificial intelligence (AI), the new Snapdragon platform scores higher by up to 25%, depending on the benchmark you’re looking at. None of them are perfect, but comparing something like MLPerf across two generations of chips does provide a good indication.
Again, most AI workloads are not sustained, and things like voice commands, face detection, or photography happen now and then. Things get more interesting with video as AI is potentially solicited much more, perhaps between 20 and 60 times per second.
In any case, it’s difficult for users to “get things done faster” if recording video and the AI is used to blur the background, for example. It’s more challenging to put a productivity number on these things.
In conclusion, the Snapdragon 8 Gen 2 platform is just as good as we hoped. The graphics performance gains are worth nothing and bode well for upcoming gaming and XR applications.
Benchmarks still don’t measure, including the brilliant links between the image processor and the neural processor, but we also don’t need to make this more complicated than it already is. Just read our Snapdragon 8 Gen 2 feature overview to better understand these critical details.
We’re looking forward to using commercial phones equipped with this SoC and reporting on how they perform.